Citation Context

...space because they are better able to capture the movement-related semantics associated with indoor entities [2]. A graph-based model for indoor space is proposed to support efficient indoor tracking =-=[12]-=-, which serves as the basis for the object management in this paper. The state definition and hash indexing method for indoor moving objects, proposed in [24] for continuous range monitoring over indo...

Citation Context

...viceID, objectID, ts, te⟩, which indicates the presence of the object within the device’s activation range during the time interval [ts, te]. The details of this pre-processing can be found elsewhere =-=[8]-=-. 2.2 Positioning Device Deployment Graph In a deployment, a subset of the devices, the so-called partitioning devices, partition the indoor space into cells in the sense that an object cannot move fr...

Citation Context

...and cj , and a loop edge {ci, ci} is mapped to the presence device(s) in cell ci. An in-depth study on deployment graphs, including relevant data structures and algorithms is available elsewhere work =-=[7]-=-. The deployment graph of Figure 1 is shown in Figure 2, where the label Di indicates the positioning device devicei. 20 D4,D4',D4'4 10 3040 50 D3,D3',D3'3D1,D1',D1'1 D2,D2',D2'2 12 13 D12 D13 D16 21 ...

Citation Context

... Although moving objects databases have been extensively studied in the last decade, the existing work is only limited within a single environment (e.g., free space [16], road network [18] and indoor =-=[21, 20]-=-). GMOD (Generic Moving Objects Database) is a database system that manages the complete movement passing through different environments, for example, walk from home to the parking lot, drive the car ...

Citation Context

... a moving object in a reader’s reading zone over a closed time period. Thus, the level of raw RFID readings is lifted by employing a pre-processing module (the details of which can be found elsewhere =-=[10]-=-) that condenses these readings into so-called appearance records. Each appearance record has the form 8ar ≡ ⟨ar-id, obj-id, reader-id, s-time, e-time⟩ where s-time and e-time are the start and end t...

Citation Context

...he entry points and are labeled as the guard de vice ei,j controlling passage from pai to paj . Observe that rooms 103A and 103B are both part of pa2, thus indicating that PAs do not necessarily have a one-to-one correspondence to the features of the space. In addition, suite 100A defines pa3 and consists of the entire space shaded with gray, which includes rooms 101, 102, 103A, and 103B. Similarly, pa5 covers the entire floor. This illus trates that PAs can be defined hierarchically. Finally, paO defines the outdoor space. To represent this constrained space we define an indoor space model [16, 15]. Indoor space models present distinguishing fea tures which perfectly match the requirements posed by our sce nario. A major feature is that those spaces are cellular (or sym bolic), i.e., they consist of a finite set of named cells or symbolic coordinates (e.g., rooms 103A and 103B) [3]. Moreover, indoor spaces may present complex topologies. Among the various topolo gies that one can specify, the most relevant are the connectivity between the indoor and outdoor spaces, as well as connectivity be tween cells (i.e., PAs). We choose to define the indoor space model as the tuple (P, E , G,...

Citation Context

...bsumes two other roles part-time lecturer and full-time lecturer, every time there is a rule specifying a condition based on role lecturer, the rule is duplicated into two new rules, corresponding to the part-time and full-time counterparts. Note that the pre-processing is trivial and incurs minimal storage overhead for moderately sized hierarchies. Fig. 2. Spatial model of protected areas in an indoor environment 2.2 Spatial Modeling We assume a bounded and discrete geographical space with respect to which access control policies are defined and enforced. Similar to the work of Jensen et al. [8], [9], we consider that the space is divided into n−1 bounded protected areas (pa), whereas the remainder of the space constitutes an additional nth area. Figure 2 illustrates a typical representation of the space, with six protected areas labeled pa1, . . . , pa6. The remainder of the space is marked as T . Observe that protected areas are logically defined based on topological relations, and do not necessarily indicate geometric properties (e.g., there is no implication that protected areas are equivalent in size or shape). Internally, protected areas are modeled as nodes in a graph, while e...

Citation Context

...or space and indoor moving objects. All the graph-based models take either rooms or doors as the nodes, and have some limitations on supporting various applications. For example, the deployment model =-=[4,5]-=- uses rooms as the nodes and sensors as the edges between rooms. It can effectively support indoor moving objects tracking and monitoring, but is not efficient when executing a KNN search on indoor mo...

Citation Context

...of the most popular variations is the Kalman filtering. Kalman filters introduced a pre-computable term, the Kalman gain, to reduce the computation load when new data arrives. Kalman filters also set a number of new constraints. The most notable constraints are the Gaussian noise and linear dynamic system assumptions. Many of the recent works are focused on relaxing these constraints [6], [7], [8]. In the graph-based approach, the topology of the graph is used to represent actual locations, and the tracking process is to categorize the detecting sensors based on their location information. In [9], for example, the rooms and the hallways are modeled as nodes in the graph, and the target is assumed to transit from node to node. Another graph-based tracking method is shown in [10]. Other graph-based methods study the overlapping of the sensing region with the aid of the known network topology [11], [12], [13]. All of these graph based algorithms depend heavily on knowledge of the network topology. The topology has to be regularly distributed and the node locations have to be known. Large targets are usually modeled by their event boundaries. Some algorithms are developed for event bounda...

Citation Context

...nnot represent the comprehensive trips above. We classify the current state-of-the-art into three categories: 1. free space [42, 53, 14, 20, 16]; 2. road (spatial) network [50, 22, 44, 18]; 3. indoor =-=[26, 25]-=-. Free space is an environment with no movement constraint. But in practice, objects usually move on a predefined set of trajectories as specified by the underlying network (road, highway, etc). The f...